![Sample our Social Sciences journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/110822/TOC-Subject-Sample-2021-Social-Sciences.jpg"})
Abstract
This study provides a primary step towards exploring whether rehabilitation efforts informed by the risk, needs, responsivity approach should be leveraged to decrease gun violence. Through the use of competing risks survival analyses, we assess the gun offense recidivism patterns of people released from prison that do (n = 1,158) and do not (n = 9,868) have gun crime conviction histories. We then explore whether gun offense recidivism increases along with actuarially based risk, how gun offense histories impact the odds of receiving community-based programming during the transition from prison to the community, and, in turn, whether programming impacts gun offending recidivism. Findings indicate that people with a history of gun offense convictions are at more than twice the hazard of committing gun offenses than similarly situated people without such histories. Predicted subhazards of recidivism and magnitudes of differences between offense history groups increase substantially as actuarially assessed risk for recidivism increases. However, predicted probabilities of receipt of community-based programming do not significantly differ between the groups with and without gun offense histories, and recidivism hazards did not meaningfully differ between those that do and do not receive community-based programs despite their gun offending histories. The results illuminate a need to expand reentry-based services towards addressing the criminogenic needs of people previously convicted of gun offenses.
Notes
1 Public Law No. 90-618, 82 Stat 1213
2 We follow similar classification strategies as Huebner, Varano, and Bynum (Citation2007) and Wallace, et al. (Citation2016) in constructing our gun offense history variable. Individuals with gun offense histories include individuals, upon their release from prison, who had previous convictions for felony weapons charges including whether individuals were in possession of guns during an offense, whether a firearm was used, shown, or threatened during an offense, and/or if a firearm was used upon a victim during an offense. Specific offenses include but are not limited to: unlawful possession of a machine gun, handgun, rifle, or shotgun; possession of a defaced firearm; possession of firearms by persons previously convicted of specified offenses. Within our data, weapons offenses are often paired with separate charges for violent offenses including, but not limited to: aggravated assault, murder, homicide, manslaughter, robbery, and/or sex crimes. Gun violence is likely apparent when these types of charges accompany one another.
3 Those that are released to parole supervision include an additional competing risk for parole revocations stemming from non-criminal rules infractions of supervision including, but not limited to: curfew violations, missed appointments with a parole officer and/or treatment staff, changing residence without permission, and/or failed urine tests.
4 Because the time to failure references dates of rearrest (that culminate in a reconviction), we resultantly refer to the outcome as hazard of rearrest when reporting results from survival analyses.
5 We constrain these analyses to the parole population because those that are unconditionally released at the end of their prison sentence are not able to participate in community-based programs offered through the parole board.
6 The use of survival analysis techniques is legion within criminological studies, particularly those that attempt to assess recidivistic behavior patterns and/or evaluate corrections and reentry programs. However, methodological and statistical challenges can present themselves when individuals can experience multiple mutually exclusive failure events during the follow-up window. Competing risks survival techniques are routinely used in medical research to attempt to mitigate, for example, potential avenues by which patients can experience mortality that is not directly related to treatment (e.g., cancer relapse, medical complications, infection, and non-medical causes of death). In this study, those attempting to reintegrate back into their community can experience rearrests associated with multiple different kinds of charges during the follow-up period (e.g., drug offenses, robberies, burglaries, property offenses, or gun crimes). Because of our interest in predicting the hazards of a cause specific event occurring during the follow-up period (rearrests for gun crimes), and that a formerly incarcerated individual can only experience one type of rearrest as their first failure event after release (a rearrest involving gun charges or any other type of rearrest that does not include such charges), the competing risks framework is well suited to our analytic scenario.
7 Variance inflation factors (VIF) for the independent variables were all at acceptable levels (e.g., the highest VIF was for the variable age at release in Model 1: VIF = 2.66, µ VIF = 1.44), as were all tolerance diagnostics (e.g., age had the lowest tolerance=.37).
8 Those with and without gun offense histories significantly differed from one another in regard to race (χ2=75.37, p≤.001).
9 Average scores and band classifications on the actuarial risk assessment instrument and the difference in proportions of reintegrating individuals being transitioned through community-based programs did not significantly differ between those who did or did not have a history of gun-related offenses.
10 The model provided for a significantly good fit for the data (LL=-1,259.11, χ2 = 109.61, p≤.001).
11 Both models were significantly good fits for the data (general rearrest model: LL=-5,335.51, χ2 = 1,657.33 (df = 12), R2=.13, p≤.001; gun-specific rearrest = LL=-725.20, χ2 = 134.00(df = 12), R2=.08, p≤.001).
12 A Cox proportional hazards test (not show, but available upon request) indicates that a history of gun-related offending was not predictive of general rearrests (rather than gun-crime-specific rearrests). This finding aligns well with prior research (see, for example, Huebner, Varano, & Bynum, Citation2007).
13 The model provides a good fit for the data (LL=-141.39, χ2 = 30.99 (df = 12), R2=.09, p≤.001).
14 The model provides a good fit for the data (LL=-647.26, χ2=109.61, p≤.001).
15 The data provided a good fit to the model (LL=-2,900.55, χ2 = 439.37, p≤.001).
16 In the interest of brevity, we do not provide effect size interpretations of these statistically significant findings.
17 We attempted to explore this nexus as a part of a series of post hoc analyses of our survival analysis strategy. However, due to several limitations we are unable to draw any meaningful conclusions. When constructing a model similar to Model 2 in Table 2 and including an interaction term for community programs receipt and gun violence history. It is unsurprising that the interaction effect resulted in a non-significant finding (SHR = 1.86; p = 0.273; 95% CI = 0.62-5.52) considering that receiving community programs was also not a significant predictor for gun offending in the model presented in Table 2. Upon plotting the predicted subhazards of the interaction term across risk scores, the results were in encouraging directions with gun violence recidivism increasing along with heightened risk and community programs recipients experiencing lower subhazards of failure when compared to those that did not receive programming. These results aligned with expectations about gun offense histories. Those that had histories of gun offending experienced heightened risks for reoffending relative to those without these histories, and those that received community programs experienced lower subhazards of failure within and across the gun history groups. However, due to the non-significant effect of the interaction term, the confidence intervals exhibited significant between-group crossover. Resultantly, these post hoc models are not presented here but are available upon request. These analyses were accomplished by using margins and marginsplot commands after running competing risks survival analyses (i.e., stcrreg) in the Stata 17.0 statistical software package. Results were ascertained while holding the other variables in the model at their current values and were substantively similar when holding covariates at their mean values.